BNBS

Bnbs (Bouncing nested Bendes solver) is an implementation of the nested Benders algorithm for solving multistage stochastic linear programming problems. It requires an LP-solver to work. Currently supported solvers are CPLEX6.5, SOPLEX 1.2.1 and GLPK 4.0 although adding an extra solver should be fairly straight forward. The name bouncing nested benders solver comes from one of the implemented sequencing protocols. In this protocol, the active stage bounces up and down in the tree. For a more thorough description of this, se the preprint article Memory consumption versus computational time in nested Benders decomposition for stochastic linear programm ing.

Usage

BNBS may be used and modified free of charge for research purposes by a student or staff member of a degree granting institution.

Requirements

In order to work, BNBS requires an LP solver, flex and bison.
NEWS 20030908: BNBS now also works with the output from stochplam, an open source modelling language for stochastic linear programming.

Results

Some results obtained with the solver for the POST test set. More results on the Test Set for Stochastic Linear Programming.

Download

You may download bnbs here, bnbs.tar.gz. Information on how to compile the program is given in the file README, included in the distribution.